LLM & reasoning architecture

Mystilink does not host models. You bring API keys; routing runs locally.

Setup: LLM providers. Issues: Model connection.

Two layers

Question → Metaphysics theory → charting / inquiry / draw
         → LLM → plan, write reading

UI	Controls
Theory selector	Metaphysics workflow
Settings → Models	Provider, model ID, fallback

Multi-provider routing

Primary model plus optional fallback when unavailable. Test connection when adding a config.

Privacy

Keys are not uploaded to Mystilink servers. See Design principles.